bioRxiv preprint doi: https://doi.org/10.1101/2020.12.08.416925; this version posted December 9, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Choi et al., 2020

Structure of Human BCCIP and Implications for Binding and Modification of Partner

Woo Suk Choi1, Bochao Liu2, Zhiyuan Shen2, Wei Yang1

1 Laboratory of Molecular Biology, NIDDK, National Institutes of Health, 9000 Rockville Pike, Building 5, Bethesda, MD 20892, USA

2 Rutgers Cancer Institute of New Jersey, Department of Radiation Oncology, Rutgers Robert Wood Johnson Medical School, 195 Little Albany Street, New Brunswick, NJ 08901

Corresponding author: Wei Yang ([email protected]) Zhiyuan Shen ([email protected])

Running title: Crystal structure of human BCCIPb Keywords: BRCA2, tumor suppressor, ribosome biogenesis, acetyltransferase Abbreviations:

1 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.08.416925; this version posted December 9, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Choi et al., 2020

Abstract

BCCIP was isolated based on its interactions with tumor suppressors BRCA2 and

. Knockdown or knockout of BCCIP causes embryonic lethality in mice. BCCIP

deficient cells exhibit impaired cell proliferation and instability. BCCIP also

plays a key role in biogenesis of ribosome 60S subunits. BCCIP is conserved from yeast

to humans, but it has no discernible sequence similarity to proteins of known structures.

Here we report two crystal structures of an N-terminal truncated human BCCIPb,

consisting of residues 61-314. Structurally BCCIP is similar to GCN5-related

acetyltransferases (GNATs) but contains different sequence motifs. Moreover, both

acetyl-CoA and substrate-binding grooves are altered in BCCIP. A large 19-residue flap

over the putative CoA binding site adopts either an open or closed conformation in BCCIP.

The substrate binding groove is significantly reduced in size and is positively charged

despite the acidic isoelectric point of BCCIP. BCCIP has potential binding sites for partner

proteins and may have enzymatic activity.

2 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.08.416925; this version posted December 9, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Choi et al., 2020

1 INTRODUCTION

BCCIP is a nuclear that was identified in based on its

interactions with tumor suppressors BRCA2 and p21 1, 2. In humans, there are two

isoforms resulting from alternative RNA splicing, BCCIPa (322 residues) and BCCIPb

(314 residues), which share the identical N-terminal 258 residues but differ in the

remaining C-terminal regions 1, 2. BCCIPa and b are also known as TOK-1a and TOK-

1b, respectively 1. BCCIPb is the conserved isoform in eukaryotes, from yeast, worms,

plants to mammals 2, while BCCIPa only exists in humans. In mouse, there is only one

BCCIP, which is ~70% identical to human BCCIPb. Either knockdown or knockout of

BCCIP in mice leads to embryonic lethality due to impaired cell proliferation 3, 4. BCCIP

deficient mouse embryo fibroblast cells exhibit increased sensitivity to DNA damage and

replication stress and increased chromosome instability, including chromosome breaks

and sister chromatid union 3. The yeast homolog of BCCIPb, known as BCP1, appears to

be involved in nuclear export and ribosome biogenesis 5, 6. Recently, both mouse BCCIP

and human BCCIPb were shown to be located in the nucleolus and required for rRNA

maturation and ribosome 60S subunit biogenesis 7, but with distinct features from the

yeast BCP1.

Despite its functional importance, structures of BCCIP-family proteins remain

elusive. Except for limited sequence similarity of ~50 residues to the Ca2+-binding domain

in calmodulin and M-calpain 2, BCCIP has no discernible to any

known protein. Using isomorphous replacement we have determined crystal structures of

a large fragment of human BCCIPb (aa 61-314) in two different crystal forms. To our

3 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.08.416925; this version posted December 9, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Choi et al., 2020

surprise, structurally BCCIP is homologous to the GCN5-related N-acetyltransferases

(GNATs) 8, which use acetyl-CoA to modify primary amines of lysine sidechains or protein

N termini, as well as aminoglycosides (antibiotics) and hormones (serotonin) 9-12. The

regions corresponding to the substrate and catalytic motifs in N-acetyltransferases (motifs

A-D), are also conserved among BCCIP-family members but are different from GNATs.

Whether BCCIP is an enzyme, and what its substrates might be are unclear. While we

were engaged in trying to figure out its structure-function relationship, a yeast BCP1

structure was reported earlier this year (with the structure coordinates on hold) 13. Here

we present crystal structures of the conserved isoform of human BCCIPb and its

outstanding binding surfaces for partner proteins and possibly for small molecules.

2 RESULTS AND DISCUSSION

2.1. Crystal structures of human BCCIP

We have determined crystal structures of N-terminal truncated BCCIPb (aa 61-314,

referred to as BCCIP hereafter) in two different space groups at resolutions of 3.06

(Native1) and 2.13 Å (Native2), respectively (Table 1). The two structures are

superimposable except for a 19-residue extended and flexible loop. BCCIP forms a single

domain of a/b fold, including a mixed seven-stranded b sheet surrounded on either side

by four and three long a helices (Fig. 1). The structure can be divided into two halves.

The N-terminal half (aa 61-185) folds into two baab hairpins with the four b strands (b1-

b4) forming an antiparallel sheet, and four a helices forming two hairpins (aA-aB, and

aC-aD) covering one face of the b sheet. Helix aC is preceded by a short helix aC’

4 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.08.416925; this version posted December 9, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Choi et al., 2020

(aa132-136). The C-terminal half (aa 186-314) folds in the order of ababba and forms a

three-stranded antiparallel b sheet (b5-7), with three a helices (aE-aG) on the side

opposite to helices aA-aD (Fig. 1a-b). The two halves are linked by helix aE and form a

contiguous b sheet by hydrogen bonds at the beginning of parallel strands b4 and b5.

These two strands diverge at the end giving the b sheet a V shape (Fig. 1b). No Ca2+-

binding module or metal ion-binding site can be found. The extended loop linking b6 and

b7, L67 (aa 269-287), has two dramatically different conformations. In Native1 L67 is

extended and interacts with a crystallographic symmetry mate reciprocally. But in Native2

L67 is folded next to aA-aB, and L67 and aAB are like two arms embracing the b sheet

(Fig. 1b). The N-terminal half followed by aE and b5 has the same sequence in the two

isoforms of human BCCIP, and only aF-b6-b7-aG differ between the two.

2.2. BCCIP is structurally similar to GNAT acetyltransferases

Structural similarity search by Dali revealed that BCCIP is similar to GCN5-related

acetyltransferases (GNATs), with substrates ranging from histones to antibiotics and

hormones. The N-terminal half of BCCIP is superimposable with these GNATs (Fig. 2a).

But in GNATs helices aC and aD are absent and b3 and b4 are linked by a b hairpin

directly or a random coil 9-12, 14. Helices aC and aD effectively close the open binding

groove for protein substrates, such as the histone tail bound to GCN5 14, and fill it with

hydrophobic residues. With its positively charged pocket (Fig. 1c) BCCIP may still be able

to bind small molecules, such as antibiotics and hormones.

5 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.08.416925; this version posted December 9, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Choi et al., 2020

The structurally most conserved region (motifs A and B) involving in acetyl-CoA

binding in GNATs is located between two halves of the structure at b4-b5 8. The

corresponding region is also conserved among BCCIP-family members, but the sequence

motifs of BCCIP (Fig. 1b) differ from GNATs 15.

The C-terminal half of BCCIP is oriented differently from GNATs, particularly loop

L5F, which binds the diphosphates of CoA in GNATs. BCCIP contains an insertion in L5F,

and without structural changes, its L5F would clash with the CoA moiety Fig. 2b). In case

substrate binding induces conformational changes in BCCIP, we tried co-crystallization

of BCCIP with 10 different acyl-CoA compounds. Instead of finding a bound substrate,

we obtained a different form of apo BCCIP crystal (Native2) (see section 3.2).

In addition, compared to GNATs, BCCIP has the following two insertions, the

extended flexible loop L67, which is open and extended in Native1 and adopts a closed

conformation in Native2 (Fig. 2c), and the C-terminal helix aG. Together with L5F, these

insertions alter the acetyl-CoA binding site significantly. Peculiarly, two Tyr residues, Y64

and Y71, are exposed

2.3. Potential binding surface for protein and small molecules

The conserved residues among BCCIP homologs are clustered into two groups.

One is around b1-aA and the loop linking b2 and b3, forming a predominantly negatively

charged convex surface (Fig. 1c). Peculiarly two conserved Tyr, Y64 and Y71, are solvent

exposed and begging for an interacting partner. The other is between b4 and b5, the V-

shaped substrate-binding groove and pocket (top and front in Fig. 1c). Aromatic residues

6 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.08.416925; this version posted December 9, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Choi et al., 2020

also abound the second cluster. Surrounding the V-shaped binding groove, two

disordered loops L23 (aa 110-119) and L5F (aa 229-244) are next to the flexible flap L67

(Fig. 1a-b), like tentacles waiting to embrace a catch. Human, mouse and yeast BCCIPs

are overall negatively charged with isoelectric point below 4.8. However, substrate

binding groove and pocket have a positively charged appearance, but they are

surrounded by an intensely negatively charged belt (Fig. 1c). These features indicate that

BCCIP binds specific protein partners and possibly a small molecule substrate in the

same location as GNATs.

3 MATERIALS AND METHODS

3.1. BCCIP expression and purification

The N-terminal 60 residues of BCCIPb were predicted to be highly disordered, and

probably prevented crystallization of the full-length BCCIPb (data not shown). The N-

terminal truncation construct of BCCIPb (aa 61-314, abbreviated as BCCIP) was cloned

into pET-28c(+) (Novagene Corp. Inc.) between NcoI and EcoRI sites, and the resulting

plasmid was transformed into BL21 (DE3) competent cells. The cells were grown in LB

medium at 37°C until OD600 reached ~0.6. After cooling down to 16°C for 30 min,

expression of BCCIP was induced by addition of 0.3 mM IPTG at 16 °C for 20 hours. The

cells were pelleted, re-suspended in buffer A (20 mM sodium phosphate (pH7.4), 500 mM

NaCl) with 40 mM imidazole, and lysed by sonication. The lysate was cleared by

centrifugation at 30,000g for 1 hour at 4°C. BCCIP was bound to an Ni2+ column and

eluted in buffer A plus 300 mM imidazole. The eluted protein was further purified using a

Hitrap Q HP column (GE Healthcare) equilibrated in buffer B (25mM Tris-HCl (pH 7.5),

7 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.08.416925; this version posted December 9, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Choi et al., 2020

100 mM NaCl, 1 mM EDTA, 0.01% IGEPAL-CA630, and 1 mM DTT) and was eluted with

a NaCl gradient from 100 to 1000 mM. As a final step, BCCIP was applied to a Hiprep

Sephacryl S-200 16/60 column (GE healthcare) equilibrated in buffer B. The BCCIP peak

was pooled and concentrated to ~15 mg/ml and stored at -80 °C in buffer B and 50%

glycerol.

3.2. Crystallization

BCCIP was buffer exchanged to 25 mM Tris-HCl (pH 7.5) and 1 mM DTT in an

Amicon Ultra-0.5 (Millipore, 10k cutoff) and concentrated to 15 mg/ml (514 µM).

Crystallization screening was performed at 4°C with JCSG core I (Qiagen) using the

sitting drop vapor diffusion method. The protein and precipitant were mixed at equal

volume (0.1 µl each). Native1 crystals were obtained in 0.2 M tri-sodium citrate and 18-

20% PEG3350. The initial crystals diffracted X-rays only to ~3.3 Å. After improving protein

purification and additions of 1 mM DTT, 100 mM NaCl and 50% glycerol in the protein

storage buffer, BCCIP crystals (Native1) became much more reproducible and diffracted

X-rays to nearly 3.0 Å.

After realizing that BCCIP is similar to GNATs, we tested co-crystallization of

BCCIP with various acyl-CoA compounds 16. Among 10 acyl-CoAs we tried, crystals

grown with 1.1 mM benzoyl(bz)-CoA with a precipitant of 0.1 M citric acid (pH 4.5) and 1-

2% PEG 6000 at 20°C diffracted X-rays to 2.13 Å. But after structural determination by

molecular replacement, the resulting electron density map showed no bz-CoA. Bz-CoA

thus functioned as an additive in crystallization, and it changed the crystal space group

with improved X-ray diffraction, to a form we termed Native2 (Table 1).

8 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.08.416925; this version posted December 9, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Choi et al., 2020

The BCCIP crystals were transferred into the respective stabilization solutions

(Native1 in 0.2 M tri-sodium citrate (pH 8.0), 25% PEG 3350, and 1 mM DTT, Native2 in

0.1 M citric acid (pH 4.5), 5% PEG 6000, and 1 mM DTT) for 30 min before freezing.

Native2 crystals cracked into smaller pieces in the stabilization solution. BCCIP crystals

were cryo-protected by 20% (Native1) or 3 % (Native2) ethylene glycol and flash-frozen

in liquid nitrogen.

3.3. Phase determination

We selected twelve commonly used Hg (mercury), Pt (platinum), Au (gold) and Pb

(lead) compounds, each of which was tested in 0.2 mM or 5 mM concentration, and

soaked into the crystals for 1 to 24 hours. We monitored crystal morphology, birefringence

and X-ray diffraction at various time intervals. Before flash-freezing in liquid nitrogen,

crystals were washed in the stabilization buffer for 15-20 second. Riso between a heavy-

atom soaked and native crystal was calculated using Scaleit implemented in CCP4i 17, 18

to evaluate whether a derivative was made. We determined the BCCIP structure with

phases obtained from six datasets of ethylmercuri-thiosalicylic acid derivative (3.15-4.2

Å, Table 1).

3.4. Data collection and structure determination

All X-ray data were collected at beamline 22ID (Native1) and 22BM (Native2 and

Hg-derivative) at the Advanced Photon Source (APS) at Argonne National Laboratory.

Each dataset was processed using HKL2000 (Native1 and Hg derivative) 19 and XDS

20 (Native2) . Native1 crystals were in P41212 space group with one molecule per

9 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.08.416925; this version posted December 9, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Choi et al., 2020

asymmetric unit, and Native2 crystals were in P21 space group with three molecules per

asymmetric unit (Table 1). For SAD (Single-wavelength Anomalous Dispersion) phasing,

6 Hg-derivative datasets were merged to enhance the anomalous signal using HKL2000

at 3.6 Å resolution (Table 1) 19, and phases were calculated using AutoSol in Phenix 21.

The electron density map was further improved by phase extension to 3.04 Å with the

native data. An initial model was built using AutoBuild in PHENIX 21. Because diffraction

of Native1 crystals was highly anisotropic, we corrected the data scaling and truncation

using Diffraction Anisotropy Server 22 to improve the map and refinement. Aided by the

secondary structure prediction by PsiPred 23 we manually traced 228 out of 254 residues

using COOT 24 and refined the BCCIP structure using PHENIX 21. (Table 1). Two loops

linking b2 to b3 (L23, aa 110-119) and b5 to aF (L5F, aa 229-244) were disordered (Fig.

1).

Native2 structure was determined by molecular replacement using Phaser in

PHENIX 21 with Native1 BCCIP structure as the search model. Model building and

refinement were carried out as described above, and non-crystallographic symmetry

constraints were applied. Except for the re-orientation of the long loop L67 (aa 269-287)

due to different crystal lattice contacts, the two BCCIP structures were superimposable

(Fig. 1).

All structure figures were prepared with PYMOL (www.pymol.org).

10 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.08.416925; this version posted December 9, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Choi et al., 2020

Table 1 Statistics of crystallographic data collection and structure refinement

Crystal Hg derivative Native1 Native2 PDB 7KYQ 7KYS

Space group P 41212 P 41 21 2 P 21 Unit cell a, b, c (Å) 114.4, 114.4, 53.9 112.7, 112.7, 56.4 60.0, 67.0, 99.5 α, β, γ (˚) 90.0, 90.0, 90.0 90.0, 90.0, 90.0 90.0, 91.0, 90.0 Resolution (Å) a 40.0 - 3.62 (3.68 - 3.62) 40.0 - 3.04 (3.09 - 3.04) 30.0 - 2.13 (2.19 – 2.13) Wavelength (Å) 1.0000 1.0000 1.0000 a Rmerge (%) 11.0 (86.7) 10.7 (98.4) 9.1 (157.9) I / σ a 22 (17) 26 (2.3) 7.8 (1.0) Completeness (%) a 100.0 (100.0) 99.8 (99.4) 98.9 (99.9) Multiplicity a 20.2 (21) 12.4 (12.1) 3.7 (3.6) Refinement Resolution (Å) a 39.8 – 3.06 29.7 – 2.13 No. reflections 6877 43893

No. reflections for Rfree 367 1001

Rwork / Rfree 0.219 / 0.251 0.209 / 0.234 No. atoms Overall 1894 5767 Protein 1850 5490 Ligand - 39 Water 44 238 Mean B-factors Overall 86.8 69.5 Protein 86.9 69.5 Ligand - 76.3 Water 84.1 69.2 R.M.S. deviations Bond lengths (Å) 0.013 0.009 Bond angles (˚) 1.539 1.262 Ramachandran Favored (%) 95.54 99.4 Allowed (%) 4.46 0.6 Outliers (%) 0 0 a Values in parenthesis are of the highest resolution shell.

11 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.08.416925; this version posted December 9, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Choi et al., 2020

FIGURES

Figure 1 Crystal structures of BCCIP. (a) Topology diagram of BCCIP. The disordered loops are indicated by dashed lines. (b) Cartoon diagrams of superimposed Native1 (semi-transparent grey and pink) and Native2 (solid rainbow colors) structure of BCCIP. Conserved residues are shown as sticks and labeled. (c) Molecular surface of BCCIP (Native2) with electrostatic potential in the same view as in panel b. The potential substrate-binding pocket and interface for protein partners are indicated.

12 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.08.416925; this version posted December 9, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Choi et al., 2020

Figure 2 Structural comparison with GCN5. (a) Superposition of the N-terminal half of BCCIP and GCN5 (PDB: 1QSN). (b) After superimposing the N-terminal half, the C- terminal half of BCCIP appears rotated by over 90° (indicated by the grey arrowhead) relative to that of GCN5. Loop L5F in BCCIP would clash with acetyl-CoA (bound to GCN5) if present. (c) The entire BCCIP structure is superimposed with GCN5 via the N-terminal half. Histone peptide bound to GCN5 is shown in a yellow (carbon) stick model.

ACKNOWLEDGEMENTS The authors thank Dr. M. Gellert for critical reading of the manuscript. The research was supported by the NIH intramural research funding to W. Y. (DK036146) and NIH grant (R01CA195612) to Z.S.

AUTHOR CONTRIBUTIONS Bochao Liu: BCCIP preparation and initial crystallization screen. Woo Suk Choi: Crystallization, structure determination, analysis and validation, and manuscript preparation. Wei Yang: structure analysis, figure preparation and writing the original draft. Zhiyuan Shen: Conceptualization; investigation, writing and editing.

CONFLICT OF INTEREST The authors declare no potential conflict of interest.

13 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.08.416925; this version posted December 9, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Choi et al., 2020

REFERENCES

1. Ono, T., et al., TOK-1, a novel p21Cip1-binding protein that cooperatively enhances p21-dependent inhibitory activity toward CDK2 kinase. J Biol Chem, 2000; 275(40): p. 31145-54. 2. Liu, J., et al., Inhibition of breast and brain cancer cell growth by BCCIPalpha, an evolutionarily conserved nuclear protein that interacts with BRCA2. Oncogene, 2001; 20(3): p. 336-45. 3. Lu, H., et al., Essential roles of BCCIP in mouse embryonic development and structural stability of . PLoS Genet, 2011; 7(9): p. e1002291. 4. Lu, H., et al., Spontaneous Development of Hepatocellular Carcinoma and B-Cell Lymphoma in Mosaic and Heterozygous Brca2 and Cdkn1a Interacting Protein Knockout Mice. Am J Pathol, 2020; 190(6): p. 1175-1187. 5. Audhya, A. and S.D. Emr, Regulation of PI4,5P2 synthesis by nuclear-cytoplasmic shuttling of the Mss4 lipid kinase. EMBO J, 2003; 22(16): p. 4223-36. 6. Li, Z., et al., Rational extension of the ribosome biogenesis pathway using network- guided genetics. PLoS Biol, 2009; 7(10): p. e1000213. 7. Ye, C., et al., BCCIP is required for nucleolar recruitment of eIF6 and 12S pre- rRNA production during 60S ribosome biogenesis. Nucleic Acids Res, 2020. 8. Marmorstein, R. and S.Y. Roth, Histone acetyltransferases: function, structure, and catalysis. Curr Opin Genet Dev, 2001; 11(2): p. 155-61. 9. Chang, Y.Y. and C.H. Hsu, Structural basis for substrate-specific acetylation of Nalpha-acetyltransferase Ard1 from Sulfolobus solfataricus. Sci Rep, 2015; 5: p. 8673. 10. Hickman, A.B., et al., The structural basis of ordered substrate binding by serotonin N-acetyltransferase: enzyme complex at 1.8 A resolution with a bisubstrate analog. Cell, 1999; 97(3): p. 361-9. 11. Wolf, E., et al., X-ray crystallographic studies of serotonin N-acetyltransferase catalysis and inhibition. J Mol Biol, 2002; 317(2): p. 215-24. 12. Stogios, P.J., et al., Structural and Biochemical Characterization of Acinetobacter spp. Aminoglycoside Acetyltransferases Highlights Functional and Evolutionary Variation among Antibiotic Resistance Enzymes. ACS Infect Dis, 2017; 3(2): p. 132-143. 13. Lin, M.H., et al., The crystal structure of protein-transporting chaperone BCP1 from Saccharomyces cerevisiae. J Struct Biol, 2020; 212(1): p. 107605. 14. Rojas, J.R., et al., Structure of Tetrahymena GCN5 bound to coenzyme A and a histone H3 peptide. Nature, 1999; 401(6748): p. 93-8. 15. Neuwald, A.F. and D. Landsman, GCN5-related histone N-acetyltransferases belong to a diverse superfamily that includes the yeast SPT10 protein. Trends Biochem Sci, 1997; 22(5): p. 154-5. 16. Zhao, S., X. Zhang, and H. Li, Beyond histone acetylation-writing and erasing histone acylations. Curr Opin Struct Biol, 2018; 53: p. 169-177. 17. Winn, M.D., et al., Overview of the CCP4 suite and current developments. Acta Crystallogr D Biol Crystallogr, 2011; 67(Pt 4): p. 235-42.

14 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.08.416925; this version posted December 9, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Choi et al., 2020

18. Howell, P.L. and G.D. Smith, Identification of heavy-atom derivatives by normal probablility methods. J. Appl. Cryst., 1992; 25: p. 81-86. 19. Otwinowski, Z. and W. Minor, Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol, 1997; 276: p. 307-26. 20. Kabsch, W., Integration, scaling, space-group assignment and post-refinement. Acta Crystallogr D Biol Crystallogr, 2010; 66(Pt 2): p. 133-44. 21. Adams, P.D., et al., PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr D Biol Crystallogr, 2010; 66(Pt 2): p. 213-21. 22. Strong, M., et al., Toward the structural genomics of complexes: crystal structure of a PE/PPE protein complex from Mycobacterium tuberculosis. Proc Natl Acad Sci U S A, 2006; 103(21): p. 8060-5. 23. Buchan, D.W.A. and D.T. Jones, The PSIPRED Protein Analysis Workbench: 20 years on. Nucleic Acids Res, 2019; 47(W1): p. W402-W407. 24. Emsley, P., et al., Features and development of Coot. Acta Crystallogr D Biol Crystallogr, 2010; 66(Pt 4): p. 486-501.

15